Muse Wonder Full: benchmark results

Meta's higher-capability Muse Wonder Full (October 12, 2026), available through paid Muse Code subscriptions with a one-million-token context. The publisher does not offer a public checkpoint download. Provider: Meta. Released 2026-10-12. Access: Interface.

Aggregate Rating 1916 ± 12, rank #17 of 10318 rated models, from 172 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA Coding Agent Index v2.0 - SWE-Atlas-QnA-AA68.82Task Resolve Rate (%)100
AA Coding Agent Index v2.1 - SWE-Atlas-QnA-AA68.82Task Resolve Rate (%)100
AutomationBench-AA - HR - Raw Objective Completion83.32Raw Objective Completion (%)100
Harvey LAB-AA13.06Hallucination-Gated All-Pass Rate (%)100
Harvey LAB-AA - Corporate M&A98.18Criterion Pass Rate (%)100
Harvey LAB-AA - Funds & Asset Management95.72Criterion Pass Rate (%)100
Harvey LAB-AA - International Trade & Sanctions96.41Criterion Pass Rate (%)100
Harvey LAB-AA - Real Estate98.06Criterion Pass Rate (%)100
Harvey LAB-AA - White Collar Defense & Investigations95.74Criterion Pass Rate (%)100
LLM Stats (DeepSWE 1.1)79Score (%)100
LLM Stats (JobBench)75.3Score (%)100
LLM Stats (Personal Superintelligence Index)59.1Score (%)100

Interactive version: theaggregate.ai/model?slug=muse-wonder-full · How It Works · Data refreshed daily, snapshot 2026-11-14.